CN-121996938-A - Regional carbon emission prediction method and device, electronic equipment and storage medium

CN121996938ACN 121996938 ACN121996938 ACN 121996938ACN-121996938-A

Abstract

The application provides a regional carbon emission prediction method, a regional carbon emission prediction device, electronic equipment and a storage medium, and relates to the field of data processing. The method comprises the steps of obtaining first carbon emission characteristics according to first historical carbon emission data of first-level administrative objects and second historical carbon emission data of second-level administrative objects governed by the first-level administrative objects in a target area, obtaining first adjacency relation characteristics according to first geographic adjacency relation data among the first-level administrative objects and second geographic adjacency relation data among the second-level administrative objects, respectively carrying out characteristic decomposition on the first historical carbon emission data and the second historical carbon emission data to obtain component characteristics on a plurality of different time scales, fusing the component characteristics with the first carbon emission characteristics to obtain second carbon emission characteristics, and inputting a prediction period, the second carbon emission characteristics and the first adjacency relation characteristics into a graph convolution network to obtain a carbon emission prediction result of the target area. The method and the device can improve the accuracy of the carbon emission prediction result.

Inventors

GU Yu
CHEN LEI
TANG XU
WANG DANLI

Assignees

内蒙古数字信息有限公司
中国科学院自动化研究所

Dates

Publication Date: 20260508
Application Date: 20251202

Claims (10)

1. A regional carbon emission prediction method, comprising: Acquiring first historical carbon emission data corresponding to first-level administrative objects in a target area, first geographic adjacent relation data among the first-level administrative objects, second historical carbon emission data corresponding to second-level administrative objects under the control of the first-level administrative objects, second geographic adjacent relation data among the second-level administrative objects and a prediction period, wherein the carbon emission data is time sequence data; obtaining a first carbon emission characteristic corresponding to the target area according to the first historical carbon emission data and the second historical carbon emission data; obtaining a first adjacency feature corresponding to the target area according to the first geographic adjacency data and the second geographic adjacency data; performing feature decomposition on the first historical carbon emission data and the second historical carbon emission data to obtain component features of the first historical carbon emission data and the second historical carbon emission data on a plurality of different time scales respectively; fusing a plurality of the component features with the first carbon emission feature to obtain a second carbon emission feature; And inputting the predicted period, the second carbon emission characteristic and the first adjacency relation characteristic into a preset graph convolution network to obtain a carbon emission predicted result of each primary administrative object in the target area in the predicted period.
2. The regional carbon emission prediction method of claim 1, wherein the obtaining the first carbon emission characteristic corresponding to the target region from the first historical carbon emission data and the second historical carbon emission data comprises: obtaining a first-level carbon emission characteristic corresponding to the target area according to the first historical carbon emission data; Obtaining secondary carbon emission characteristics corresponding to each primary administrative object according to the historical carbon emission data corresponding to the secondary administrative object governed by each primary administrative object in the second historical carbon emission data; and carrying out enhancement treatment on the primary carbon emission characteristic according to the secondary carbon emission characteristic to obtain a first carbon emission characteristic corresponding to the target region.
3. The regional carbon emission prediction method according to claim 2, wherein the enhancing the primary carbon emission characteristic according to the secondary carbon emission characteristic to obtain the first carbon emission characteristic corresponding to the target region includes: for each primary administrative object in the target area, determining the average value of the corresponding secondary carbon emission characteristics of the primary administrative object; and fusing the average value corresponding to each first-level administrative object with the corresponding characteristic of the corresponding first-level administrative object in the first-level carbon emission characteristics to obtain the first carbon emission characteristics.
4. The regional carbon emission prediction method according to claim 1, wherein the obtaining the first adjacency feature corresponding to the target region from the first and second geographical adjacency data includes: obtaining first-level adjacency characteristics corresponding to the target area according to the first geographic adjacency data; Obtaining a secondary adjacency feature corresponding to the target area according to the second geographic adjacency data; And carrying out enhancement processing on the first-level adjacent relation characteristic according to the second-level adjacent relation characteristic to obtain the first adjacent relation characteristic.
5. The regional carbon emission prediction method of claim 4, wherein the enhancing the primary adjacency feature according to the secondary adjacency feature to obtain the first adjacency feature comprises: Performing aggregation treatment on the secondary adjacency feature to obtain an aggregated secondary adjacency feature; And fusing the aggregated secondary adjacency feature with the primary adjacency feature to obtain the first adjacency feature.
6. The regional carbon emission prediction method of claim 1, wherein performing feature decomposition on the first historical carbon emission data and the second historical carbon emission data to obtain component features of the first historical carbon emission data and the second historical carbon emission data on a plurality of different time scales each comprises: performing feature decomposition on the first historical carbon emission data through an aggregate empirical mode decomposition method to obtain at least one first eigenmode function component and a first residual component, and obtaining component features of the first historical carbon emission data on a plurality of different time scales according to the at least one first eigenmode function component and the first residual component; And performing feature decomposition on the second historical carbon emission data through the aggregate empirical mode decomposition method to obtain at least one second eigenmode function component and a second residual component, and obtaining component features of the second historical carbon emission data on a plurality of different time scales according to the at least one second eigenmode function component and the second residual component.
7. The method for predicting regional carbon emissions according to claim 1, wherein the graph packing network comprises full-connected layers and a predetermined number of graph packing layers, wherein the step of inputting the predicted period, the second carbon emission characteristic and the first adjacency feature into the predetermined graph packing network to obtain a carbon emission prediction result of each first-level administrative object in the target region in the predicted period comprises the steps of: Performing feature extraction on the second carbon emission feature and the first adjacent relation feature through the graph convolution layer with the preset number to obtain a convolution feature; processing the convolution characteristics and the prediction period through the full connection layer to obtain a carbon emission prediction result of each first-level administrative object in the target area in the prediction period; Wherein each of the graph convolution layers performs feature extraction using the following formulas (1) - (3), and the carbon emission prediction result is represented by the following formula (4): (1) (2) (3) (4) Wherein, the , To represent a matrix of second carbon emission characteristics, To represent the matrix of first adjacency features, Is a matrix of units which is a matrix of units, Is the first The input feature matrix of the layer is, In the form of a degree matrix, Is the first A matrix of the learnable parameters of the layer, In order to activate the function, In order to be able to set the number in advance, As the weight of the full connection layer, Is the bias of the full link layer.
8. A regional carbon emission prediction apparatus, comprising: The first acquisition module is used for acquiring first historical carbon emission data corresponding to the first-level administrative objects in the target area, first geographic adjacent relation data among the first-level administrative objects, second historical carbon emission data corresponding to the second-level administrative objects under the jurisdiction of the first-level administrative objects, second geographic adjacent relation data among the second-level administrative objects and a prediction period, wherein the carbon emission data is time sequence data; The second acquisition module is used for acquiring a first carbon emission characteristic corresponding to the target area according to the first historical carbon emission data and the second historical carbon emission data; The third acquisition module is used for acquiring a first adjacent relation characteristic corresponding to the target area according to the first geographic adjacent relation data and the second geographic adjacent relation data; the characteristic decomposition module is used for carrying out characteristic decomposition on the first historical carbon emission data and the second historical carbon emission data to obtain component characteristics of the first historical carbon emission data and the second historical carbon emission data on a plurality of different time scales; The feature fusion module is used for fusing the plurality of component features with the first carbon emission features to obtain second carbon emission features; And a fourth obtaining module, configured to input the predicted period, the second carbon emission characteristic, and the first adjacency relation characteristic into a preset graph convolution network, to obtain a carbon emission predicted result of each first-level administrative object in the target area in the predicted period.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and running on the processor, wherein the processor implements a regional carbon emission prediction method as claimed in any one of claims 1 to 7 when the computer program is executed by the processor.
10. A non-transitory computer readable storage medium having stored thereon a computer program, wherein the computer program when executed by a processor implements a regional carbon emission prediction method as claimed in any one of claims 1 to 7.

Description

Regional carbon emission prediction method and device, electronic equipment and storage medium Technical Field The present application relates to the field of data processing, and in particular, to a regional carbon emission prediction method, a regional carbon emission prediction device, an electronic device, and a storage medium. Background Global climate change has become the biggest non-traditional safety challenge for human development, and carbon peak and carbon neutralization are significant in actively coping with this challenge. The largest carbon emission country in the world is always in focus of reducing carbon emission, so that the current carbon emission situation in China is clarified, the main factors influencing carbon emission are analyzed, the future carbon emission trend is predicted, and the method has profound significance for carbon emission reduction in China. Currently, some carbon emission analysis techniques have advanced in practical applications, such as regional carbon emission monitoring, industry emission contribution assessment, and the like. However, for fine-grained prediction of carbon emissions, especially complex relational modeling and analysis techniques in combination with multidimensional data, large-scale applications have not been realized, mainly because the performance of existing models is difficult to meet the actual demands. Carbon emission predictions typically require modeling from two dimensions, spatial and time series, to identify inter-regional carbon emission interactions and their timing variations. The existing mainstream method generally adopts the same processing mode for the two features, for example, spatial features and temporal features are directly spliced and then used as input, but the method is easy to lead to the introduction of redundant information. In fact, according to the difference between prediction areas, not only the complex dynamic change of the time sequence features, but also the spatial correlation between the areas need to be considered, so that the indiscriminate processing method not only increases the calculation load, but also can prevent effective mining of important information and rules in the carbon emission data, and reduces the accuracy and efficiency of prediction. In addition, in the existing method, a complex fusion strategy is mostly adopted, namely spatial features and temporal features are directly and uniformly coded, joint feature characterization is generated, and then carbon emission values are predicted through the joint features. However, in a practical scenario, the contribution of the data features of different samples to the prediction task is non-uniform, and for the prediction of carbon emission in certain areas, high prediction accuracy may be obtained only by using time features, and joint feature characterization is forcedly used, but redundant information is easily introduced, so that mutual interference between different features is caused, and the prediction accuracy is reduced. Disclosure of Invention The application provides a regional carbon emission prediction method, a regional carbon emission prediction device, electronic equipment and a storage medium, which are used for solving the defects that the spatial characteristics and the temporal characteristics are processed in the same way and a complex fusion strategy is adopted for the spatial characteristics and the temporal characteristics in the prior art, and can obviously improve the accuracy of carbon emission prediction. The application provides a regional carbon emission prediction method, which comprises the following steps: Acquiring first historical carbon emission data corresponding to first-level administrative objects in a target area, first geographic adjacent relation data among the first-level administrative objects, second historical carbon emission data corresponding to second-level administrative objects under the control of the first-level administrative objects, second geographic adjacent relation data among the second-level administrative objects and a prediction period, wherein the carbon emission data is time sequence data; obtaining a first carbon emission characteristic corresponding to the target area according to the first historical carbon emission data and the second historical carbon emission data; obtaining a first adjacency feature corresponding to the target area according to the first geographic adjacency data and the second geographic adjacency data; performing feature decomposition on the first historical carbon emission data and the second historical carbon emission data to obtain component features of the first historical carbon emission data and the second historical carbon emission data on a plurality of different time scales respectively; fusing a plurality of the component features with the first carbon emission feature to obtain a second carbon emission feature; And inputting the predicted perio